Stromal TRIM28-associated signaling pathway modulation within the colorectal cancer microenvironment

The study was approved by Ethics (Medical) Research Committee at Beaumont Hospital, Dublin, Ireland. Informed consent was obtained from all patients. Patients with a history of cancer and neo-adjuvant treatment were excluded. A total of 19 cases with late-stage (stages III and IV) colorectal cancer (CRC) were investigated. All patients were diagnosed with CRC between 2012 and 2013. The median age of the patients at the time of diagnosis was 67 (range 47–88 years) with 10 male and 9 female patients. In total, 15 patients had colonic carcinoma, whilst 4 had rectal carcinomas. Clinical and pathological parameters of the patient cohort are shown in Table 1.

Both, fresh-frozen and formalin-fixed and paraffin embedded (FFPE) tissue samples were obtained for each case. FFPE tissues were used for immunohistochemical analysis, while fresh-frozen tissues were used for laser capture microdissection (LCM) and reverse phase protein array (RPPA) analysis. A pathologist identified and collected areas of invasive carcinoma from the tumor mass for fresh-frozen preservation and FFPE. Local resection and standard fixation protocols were carried out in all cases. Each FFPE block was sectioned and stained with hematoxylin and eosin (H&E) and graded by a consultant pathologist to confirm pathological stage and grade of the tumors. All fresh tissue samples for RPPA analysis were snap-frozen in liquid nitrogen and processed uniformly and rapidly to ensure preservation of molecular endpoints. The time from removal of a colectomy specimen to snap-freezing of samples was < 20 mins. Fresh-frozen tissue samples were stored at − 80 °C.

Immunohistochemistry (IHC) and scoring were carried out using an anti-TRIM28 rabbit monoclonal antibody (mAb) (C42G12, Cell Signaling Technology Inc., Danvers, MA, USA), as previously described [18]. The degree of nuclear TRIM28 staining was evaluated for epithelial and stromal tissue separately and scored as follows: absence of staining (score = 0), weak staining (score = 1+), moderate staining (score = 2+) and strong staining (score = 3+). TRIM28 epithelium to stroma ratios were computed based on a Protein Expression Ratio method using IHC scores as previously described [18]. Briefly, high TRIM28 Expression Ratio was defined as at least 2 units of difference in staining intensity (e.g. epithelium strong [3+] and stroma weak [1+], or epithelium moderate [2+] and stroma negative [0]). Low TRIM28 Expression Ratio was defined as 1 or 0 units of difference in staining intensity (e.g. epithelium moderate [2+] and stroma weak [1+], or epithelium weak [1+] and stroma weak [1+]). A previous study in our lab has shown that the inter-observer variability of IHC scoring is as low as 7% [22]. In cases where there were discrepancies between the scorers, a consensus was reached after a joint review using a multi-headed microscope.

Laser capture microdissection (LCM) was performed to isolate separate populations of epithelial and stromal cells for cell signaling analysis as described previously [23‐25]. Briefly, consecutive 8 µm thick frozen tissue sections were cut using a cryostat for each sample and mounted on glass slides. Then, using an infrared-based laser capture system (ArcturusXT, Applied Biosystems, San Francisco, CA, USA), approximately 20,000 epithelial and stromal cells were removed for each frozen tissue sample. To account for possible heterogeneity within the tissue sample, multiple separate areas within each patient sample were micro dissected and no attempt was made to target specific regions within the tumor. Tissue processing and preparation of tissue lysates have been described previously [26, 27].

Reverse phase protein arrays (RPPA) were generated as previously described [25, 28, 29]. Each array contained epithelial and stromal cell lysates for all 19 cases and each lysate was printed in a 2-fold dilution curve representing undiluted lysate, 1:2, 1:4 and 1:8 dilutions. Control lysates were printed in a twofold dilution curve. All RPPAs were incubated for 2 h at 80 °C to allow fixation and then stored with desiccant at − 20 °C. Quality control samples were printed on the RPPA to ensure protein deposition and immunostaining reactivity [30]. These included A431 cell lines (± EGF stimulation; BD Pharmingen, San Diego, CA, USA) and a pooled sample of CRC cases.

RPPA slides were blocked (I-Block, Applied Biosystems) for 2 h prior to immunostaining. Immunostaining was conducted on a Dako Autostainer (Catalyzed Signal Amplification (CSA) kit, Dako). Each slide was incubated with a single primary antibody at room temperature for 30 min. The negative control slide was incubated with antibody diluent (Dako) instead of a primary antibody. For normalization purposes, a slide was incubated with anti-ssDNA antibody (1:15,000; IBL International GmbH). The secondary antibodies were goat anti-rabbit (1:10,000; Vector Laboratories), or rabbit anti-mouse IgG (1:10; Dako) depending on the host species of the primary antibody. Amplification was achieved using horseradish peroxidase-mediated biotinyl tyramide with chromogenic detection (diaminobenzidine; Dako). In order to determine the total protein concentration in each sample, two RPPA slides were stained with Sypro Ruby Protein Blot Stain (Molecular Probes, Eugene, OR, USA) and visualized with NovaRay Image Acquisition Software (Alpha Innotech, San Leandro, CA, USA).

In total, 30 primary antibodies specific to known signaling endpoints were used to measure phosphorylation and protein levels using RPPAs (see Additional file 1). In an effort to elucidate the underlying proteomic mechanisms responsible for the previously reported significantly different 5-year survival between TRIM28 high and low ratio patients [18], many endpoints related to known catabolic pathways essential for organismal homeostasis such as autophagy and apoptosis were chosen. Several endpoints that are known modulators of TRIM28 function were also included.

RPPA slides were scanned on a flatbed scanner (UMAX 2100XL; white balance 255, black 0, middle tone 1.37, 600 dpi, 14 bit) and spot intensity was analyzed using the Image Quant v5.2 software package (Molecular Dynamics). To normalize data, the relative intensity for each protein spot was divided by the ssDNA relative intensity for the corresponding spot [31] and data reduction was carried out using a VBA Excel macro, RPPA Analysis Suite [32]. One case did not meet the quality criteria as the spot intensities were not significantly different from the negative control and therefore was excluded from proteomic network analysis. All staining intensities are listed in Additional file 2: Table S2.

Proteomic network analysis was performed as previously shown [25, 33]. Spearman ρ correlation analysis with ρ ≥ 0.75 and p ≤ 0.01 was used to build proteomic network graphs (Gephi 0.9, The Gephi Consortium, Paris, France, http://​www.​gephi.​org). Proteomic networks were drafted based on 18 CRC cases (one excluded) and classified into two groups based on their epithelial to stromal TRIM28 expression ratio score; TRIM28 low ratio group (n = 9) representing good patient prognosis and TRIM28 high ratio group (n = 9) representing poor patient prognosis [18]. In order to identify the differences in the proteomic signature of the cohorts, significant correlations that were common in both TRIM28 ratio groups in the same tissue compartment (epithelial or stromal) were excluded and only exclusive significant correlations were used to construct the proteomic network graphs. For example, correlations in the epithelial compartment that were common in both TRIM28 high and low ratio cohorts, were excluded and only correlations that were unique to the TRIM28 high or low ratio groups were used to construct the proteomic network graphs for each group.

Nodes in the network represent proteins quantified by RPPA; the bigger the node, the more significant correlations relative to that protein. Each line connecting 2 nodes represents a significant correlation between the nodes; the thicker the line, the higher the Spearman ρ correlation between those two proteins. Proteins within the network are grouped into various subgroups on the basis of Spearman ρ values and the number of connections among a group of nodes. Strongly correlated nodes are represented close to each other and with the same color, with each color representing a different subgroup. The number of subgroups within each proteomic network is determined by the software.

All tests were analyzed using SPSS 21.0 software (SPSS, Chicago, IL, USA) and JMP 5.1.2 (SAS Institute Inc., Cary, NC, USA). The Spearman rank correlation coefficient, ρ, was calculated for each protein pair in the RPPA cohort; ρ ≥ 0.75 with p ≤ 0.01 was considered significant. An independent-samples t test was used to compare the expression levels of each endpoint between groups, the findings were considered statistically significant at p < 0.05.

We have previously shown that the assessment of TRIM28 expression levels, when it encompasses epithelial and stromal tissue compartments, can serve as an independent marker of relapse and patient survival in colorectal cancer [34]. In order to assess stromal and epithelial TRIM28 levels in this cohort, we separately scored TRIM28 in epithelial carcinoma cells and stromal fibroblasts in 19 FFPE tissues using IHC. Strong epithelial staining (3+) for TRIM28 was found in 8 cases, moderate staining (2+) in 9 and weak staining (1+) in 2. Cells in the stromal compartment showed moderate TRIM28 staining (2+) in 2 cases, weak staining (1+) in 12 cases and absence of staining (0) in 5 cases (Fig. 1).

The tissue-compartment protein expression ratio method developed for reciprocal tumor microenvironment assessment [34], allowed us to divide the 19-case CRC cohort in 2 groups; the TRIM28 high epithelium: stroma ratio group, which is mainly characterized by low levels of TRIM28 in tumor stroma (Fig. 1a) and the low epithelium: stroma ratio group, which shows more similar TRIM28 expression levels in both, epithelial and stromal compartments (Fig. 1b). Additional examples for epithelial to stromal expression variations and their assigned ratios are shown (Additional file 3). Using the IHC scoring data, the ratio method resulted in 10 cases being assigned to the TRIM28 high ratio group and 9 cases to the low ratio group (Fig. 1c). In the high ratio group, one case had high epithelial TRIM28 levels and absence of TRIM28 in stroma expression, 6 cases had high levels of epithelial TRIM28 and weak staining in the stroma and 3 cases had moderate levels of TRIM28 in the tumor epithelium with an absence of staining for TRIM28 in stroma (Fig. 1c). Most cases in the TRIM28 low ratio group had weak to moderate expression in both, epithelium and stroma. One case had strong epithelial expression and another case presented with absence of stromal TRIM28 staining. The distribution of IHC scores allotted to both TRIM28 expression ratio groups is shown in Fig. 1c.

We next sought to understand the underlying molecular interplay associated with TRIM28 expression ratios and signaling pathways linked to cancer. RPPAs were immunostained against 30 signaling endpoints and the resulting quantitative data were used for proteomic network analyses in 9 TRIM28 high ratio and 9 TRIM28 low ratio cases (see Additional file 2). The Spearman’s rank-order correlation analysis revealed that 55 of 435 possible protein pair correlations were shared (ρ ≥ 0.75, p ≤ 0.01) between stroma of TRIM28 high and low ratio cases (see Additional file 4).

In addition, the high ratio TRIM28 stromal proteomic network revealed a moderate number of 76 exclusive signaling endpoints associations, which were not present in TRIM28 low ratio stoma (see Additional file 4). As shown in Fig. 2a, the stromal high ratio network is divided into three sub-networks. The proteomic network is dominated by nodes linked to tumor progression, including acetyl-CoA, LC3B, COX-2 and Survivin as prominent endpoints.

The low ratio TRIM28 stromal network showed an overall higher number of exclusive signaling endpoints with 81 significant associations when compared to high ratio TRIM28 stoma (see Additional file 4). As shown in Fig. 2b, the network is also divided in three sub-groups, these however are considerably different to the high ratio networks and are dominated by endpoints linked to apoptosis, such as the activated caspases 3 and 7 and PP2A. The association with apoptotic proteins is not seen in the TRIM28 high ratio stromal network and shows that the proteomic architecture of the stromal tissue compartment is significantly different in TRIM28 high and low ratio patients and hence may account for differences in patient outcomes seen in both colorectal cancer groups.

In order to obtain a complete picture of potential molecular tumor microenvironment characteristics, we performed an equivalent Spearman rank correlation mapping for the epithelial compartments of the patient cohort. The 30-endpoint Spearman rank analysis revealed that 87 of 435 possible protein pair correlations were shared between epithelia of TRIM28 high and low ratio cases (see Additional file 5), indicating more similar signaling patterns between both groups when compared to the stroma counterpart.

Fifty-five protein pairs correlated (ρ ≥ 0.75, p ≤ 0.01) exclusively in the epithelium of TRIM28 high ratio cases and were entirely absent in TRIM28 low ratio cases (see Additional file 5). Detailed graphical representation of those high ratio-exclusive associations of protein expression and phosphorylation in epithelial compartments is shown in Fig. 3a. The TRIM28 high ratio epithelial proteomic network is associated with several pro-survival endpoints including cell–cell adhesion (E-cadherin) and autophagy (AKT, Survivin).

The proteomic network analysis of tumor epithelium in the TRIM28 low ratio group revealed statistically significant protein pair correlations in 69 signaling endpoints (ρ ≥ 0.75, p ≤ 0.01) (see Additional file 5). Those pairs were exclusive to the epithelium of TRIM28 low ratio cases and were absent in TRIM28 low ratio epithelium (Fig. 3b). The proteomic network is dominated by pro-apoptotic proteins represented, among others, by activated caspase-7 and the anti-apoptotic regulator BCL-2. Those nodes and protein correlations are noticeably diminished in the high ratio epithelial proteomic network. In addition, the proto-oncogene Lck, another prominent node in the TRIM28 low ratio epithelial network, was also identified as a dominant node in TRIM28 low ratio stromal network (Figs. 2b, 3b).

We next compared expression levels of 30 signaling proteins between tumor epithelial and stromal compartments in either TRIM28 high or low ratio cases. Our analysis revealed that expression levels of MDM2 are significantly lower in stroma of TRIM28 high ratio cases when compared to corresponding epithelial compartments (p = 0.010, Fig. 4a). No statistically significant differences in MDM2 levels were found in low ratio cases (p = 0.923, Fig. 4a). In contrast, stromal expression levels of COX-2 were significantly elevated in high ratio cases when compared to tumor epithelium (p = 0.002, Fig. 4b). There are no significant differences in COX-2 intensities between epithelium and stroma of TRIM28 low ratio cases (p = 0.465, Fig. 4b). These findings further substantiate a molecular disparity between TRIM28 high and low ratio cases. Our earlier findings, for instance, show that TRIM28 may interact with MDM2 to influence the levels of the tumor suppressor p53 and this may account for significantly poorer 5-year overall survival and 5-year ‘recurrence-free’ survival seen in TRIM28 high ratio colorectal cancer patients [18].

The expression levels of all investigated signaling endpoints were then analyzed across epithelial and stromal data regardless of their TRIM28 ratio scores. We found significantly less activated caspase-3 and caspase-7 expressed in the stroma of TRIM28 high ratio patients compared to stroma of low ratio patients (p = 0.041, Fig. 5a; p = 0.036, Fig. 5b). Activated caspases are the main executioners of apoptosis in the cell. Since the reduction of stromal apoptosis was previously shown to be an independent prognostic factor for poorer overall survival and disease-recurrence in CRC [35], the lower levels of activated caspases seen in the stroma may be another contributing factor to the poorer outcomes in colorectal cancer patients with TRIM28 high ratios.